DarkSca's picture
Update README.md
de5fdff verified
---
license: mit
tags:
- bitnet
- lora
- ternary
- trillim
- cpu-inference
base_model: microsoft/bitnet-b1.58-2B-4T-bf16
---
# BitNet-Search-LoRA-TRNQ
Ternary-quantized LoRA adapter for [Trillim/BitNet-TRNQ](https://huggingface.co/Trillim/BitNet-TRNQ) that teaches the model to search. Generates search queries in `<search>...</search>` XML tags. Need to use Trillim (https://trillim.com) with a search harness to respond to the model's search calls and improve results.
This adapter runs entirely on CPU — no GPU required.
## Adapter Details
| | |
|---|---|
| **Type** | LoRA adapter |
| **Style** | Search tool call |
| **Architecture** | BitNet (BitNetForCausalLM) |
| **Quantization** | Ternary ({-1, 0, 1}) |
| **Platforms** | x86_64, aarch64 |
| **Base model** | [Trillim/BitNet-TRNQ](https://huggingface.co/Trillim/BitNet-TRNQ) |
| **Source model** | [microsoft/bitnet-b1.58-2B-4T-bf16](https://huggingface.co/microsoft/bitnet-b1.58-2B-4T-bf16) |
| **License** | MIT |
## Usage
```bash
pip install trillim
trillim pull Trillim/BitNet-TRNQ
trillim pull Trillim/BitNet-Search-LoRA-TRNQ
trillim chat Trillim/BitNet-TRNQ Trillim/BitNet-Search-LoRA-TRNQ
```
This starts an interactive CLI chat.
## What's in this repo
| File | Description |
|---|---|
| `qmodel.lora` | Ternary-quantized LoRA weights in Trillim format |
| `tokenizer.json` | Tokenizer |
| `tokenizer_config.json` | Tokenizer configuration |
| `chat_template.jinja` | Chat template |
| `trillim_config.json` | Trillim metadata |
## License
This adapter is released under the [MIT License](https://opensource.org/licenses/MIT), following the license of the source model.